EPSRC Research Fellowship in Artificial Intelligence - Monte Carlo Tree Search

[Peter Cowling]

EPSRC Research Fellowship in Artificial Intelligence (AI) Monte Carlo
Tree Search for Games and other Decision Tree Applications with
Uncertainty and Hidden Information
University of Bradford - School of Computing, Informatics and Media
Supervisor: Professor Peter Cowling (www.scim.brad.ac.uk/~picowlin)
Financial Information: Grade 8: £31,671 - £35,646 per annum. The post
duration is 3 years.
Start Date: 1st October 2010 (or later by agreement).
Deadline for Applications: 31st July 2010.

Background: The desire for better Artificial Intelligence (AI) in
board games, card games and, more recently, video games have been a
constant driver for AI research. There is considerable current
excitement over the potential of Monte Carlo Tree Search algorithms
(particularly UCT (http://senseis.xmp.net/?UCT)), which have produced
professional-level computer Go players in the past year, a significant
step towards addressing one of the major outstanding research
questions of Artificial Intelligence (http://oase.nutn.edu.tw/
FUZZ_IEEE_2009/result.htm). We have received substantial funding
from EPSRC to investigate the full potential of Monte Carlo Tree
Search (http://gow.epsrc.ac.uk/ViewGrant.aspx?GrantRef=EP/H049061/1),
in a collaboration between the University of Bradford, Imperial
College London, University of Essex, University of Reykjavik
(Iceland), AI Factory Ltd., Introversion Software Ltd. and Nestorgames
Ltd.

The Project: For the majority of problems where game/decision tree
models are appropriate, particularly board and card games, real-time
video games and other sequential decision problems, the tree is
nondeterministic, with both hidden information and randomness. The
Research Fellow will build a conceptual and a software framework for
handling uncertain decision tree nodes within Monte Carlo Tree Search,
particularly looking at methods which simplify and determinise random
outcomes and make best use of available CPU time to produce strong AI
game players/decision tools. Part of this investigation will be to
consider ensemble and multi-level approaches. We will be investigating
card and board games (and the fellow's preference will be important in
determining the actual games to be studied) as well as related
problems in General Game Playing (see http://games.stanford.edu/) and
from decision support areas such as discrete event simulation.

The Research Fellow: The Research Fellow will work closely with Prof
Peter Cowling and two PhD students working in this area (one in post
and one to be appointed), as well as maintaining regular collaborative
contact with other members of the consortium. Research into Monte
Carlo Tree Search is a new area, where the Fellow can make a
pioneering impact and get to grips will all work to date in a
relatively short time. Theoretical and experimental research skills
are needed.
Good programming skills are required, preferably in a range of
languages such as C#/C++/Python as well as a PhD in an area related to
artificial intelligence, discrete mathematics and the design of
algorithms and data structures. If the PhD area, or other experience,
encompasses search within trees or other combinatorial objects then
this is better still. An enthusiasm for games and a knowledge of
parallel programming techniques (including GPU approaches) is also
highly desirable.

The University, School and City: The School of Computing Informatics
and Media at the University of Bradford has a strong research focus,
as evidenced by our population of over 100 PhD students and the fact
that 45% of our research was rated "Internationally Leading" or "World
Class" by the 2008 Research Assessment Exercise (and all was
recognised nationally or internationally). In 2010, all research staff
and students were relocated to purpose-built facilities within the
University's city-centre campus. One of the UK's largest northern
cities, set amongst some of the most spectacular countryside in
England and yet one of the most affordable places to live in Britain,
Bradford's wonderful paradoxes create a living and working environment
that is hard to beat. Bradford also has excellent transport links via
Leeds/Bradford airport to London, Europe and the rest of the world,
and via excellent UK road and rail connections.

Academic Requirements: A good PhD in an area as above, or equivalent
experience. Candidates whose first language is not English need to
have IELTS 7.0 or equivalent.

For informal enquiries and further information: Contact Prof Peter
Cowling (email: p.i.c...@bradford.ac.uk, tel.: +44 (0) 1274
234005).

To Apply: See http://www.brad.ac.uk/human-resources/jobs-at-bradford/
(job ref RCI1273) for details of how to apply. All applicants are
encouraged to contact Peter Cowling informally if they are considering
making an application. Note the deadline (31st July 2010).